The load_persisted_configs parameter now controls whether persisted
runtime changes are loaded on startup. The changes are still generated as
it persists the current state of MaxScale making problem analysis easier.
The new `force=yes` option closes all connections to the server that is
being put into maintenance mode. This will immediately close all open
connections to the server without allowing results to return.
That URL will now return information about the statements in
the query classifier cache. The information is collected using
the same map in a serial manner from all routing workers (that
each have their own cache). Since all caches will contains the
same statements, collecting the information in a serial manner
means that the overall memory consumption will be lower than
what it would be if the information was collected in parallel.
With the addition of SO_REUSEPORT support, it is no longer possible to
rely on the network stack to prevent multiple listeners from listening on
the same port. Without explicitly checking for the ports it would be
possible for two listeners from two different services to listen on the
same port in which case the service would be almost randomly chosen.
If SO_REUSEPORT is available and the kernel supports it, listeners will
now listen on separate file descriptors. This removes the need for
cross-worker communication when in normal operation which should make
MaxScale scale better.
By storing the file descriptor inside a worker-local variable, it is
possible to handle both unique file descriptors (created with
SO_REUSEPORT) and shared file descriptors with the same code. The way in
which the file descriptor is stored in the rworker_local object determines
the way the listener behaves.
Increasing counter sizes from int to long for averages.
Rename random functions to end with _co instead of _exclusive to
indicate range [close, open[, and to allow future suffixes oc, cc and oo.
By passing strings instead of const char pointers to the task control
functions, we can safely make copies of them knowing that the contents
won't disappear.
Storing all the runtime errors makes it possible to return all of them
them via the REST API. MaxAdmin will still only show the latest error but
MaxCtrl will now show all errors if more than one error occurs.
By using a std::vector to store the keys, the cost of key lookup goes down
in comparison with std::unordered_map. The downside is the increase in
memory use in certain situations but given the fact that the intended
purpose of worker-local data is to store global data, this is unlikely to
have observable negative side-effects. Depending on the implementations of
std::vector and std::unordered_map, the use of std::vector could even
result in a smaller memory footprint.
Added an overload to execute_concurrently that takes an std::function as a
parameter and added a const version of operator* for rworker_local. Also
removed the std::move of the return value in rworker_local::values as it
can prevent RVO from taking place.
Some SQL clients may default to a different authentication plugin than
"mysql_native_password". Since this is the only one supported by MySQL-
authenticator, the client is instructed to swap its plugin.
The functions are now in MonitorServer. Disk space can only be checked
during specific ticks. If a server misses a tick (e.g. is down) it will
be checked after disk_space_check_interval has passed.
The rank can now only be used to define two groups of servers: primary and
secondary servers. This limits the exposure and reduces the number of
possibilities that can arise from the use of this parameter thus making it
more predictable.
Although the default value is the maximum value of a signed 32-bit
integer, the value is stored as a 64-bit integer. The integer type
conversion functions return 64-bit values so storing it as one makes
sense.
Currently values higher than the default are allowed but the accepted
range of input should be restricted in the future.
Previously, runtime monitor modifications could directly alter monitor fields,
which could leave the text-form parameters and reality out-of-sync. Also,
the configure-function was not called for the entire monitor-object, only the
module-implementation.
Now, all modifications go through the overridden configure-function, which calls the
base-class function. As most configuration changes are given in text-form, this
removes the need for specific setters. The only exceptions are the server add/remove
operations, which must modify the text-form serverlist.
The protocol should not track the session state as the parsing is quite
expensive with the current code. This change is a workaround that enables
the parsing only when required. A proper way to handle this would be to do
all the response processing in one place thus avoiding the duplication of
work.
